AI supply-chain thesis — mapping bottlenecks, focus companies, and supply-chain exposure for investors.
**Bottleneck theme:** HBM / Packaging
**Focus:** $ONTO — ONTO INNOVATION INC.
Onto Innovation is one of two specialty advanced-packaging metrology vendors (with Camtek $CAMT) that has emerged as a critical-tool supplier to HBM stacking and CoWoS-class advanced packaging. The 3Di optical metrology platform is sole-qualified at two of the three leading HBM customers, addressing the binding inspection problem in HBM4: bump pitch is shrinking, stack count is climbing, and the cost of a defect-induced field failure on a $5k+ HBM stack is enormous. Onto's installed base on HBM, CoWoS, and 3D NAND lines provides recurring service and consumables revenue layered on top of the tool sales.
The investment case is mid-cycle compounding plus an HBM-specific content tailwind: every HBM4 line that qualifies adds $100M+ of opportunity, and the CoWoS-L expansion at TSMC adds further per-wafer content. The bear case is the duopoly with Camtek (price competition possible), KLA's umbrella position above both vendors, and customer concentration in three HBM stackers. Pair with $CAMT for diversified HBM metrology exposure.
HBM — High Bandwidth Memory — 3D-stacked DRAM (HBM2E/HBM3/HBM3E/HBM4) connected via through-silicon vias, delivering 1+ TB/s of bandwidth per stack. Co-packaged with GPUs, TPUs, and custom AI accelerators for datacenter AI training/inference and HPC workloads.
